Today, natural language processing, machine vision, and AI and ML functions find more applications in practice. Due to the fact that humans are not able to generate structured information, they are not able to apply pure mathematics thoroughly to models.

For this reason, we use software that generates structured data. Most models (including psychological, language, sociological, etc.) are based on the human perseverance of experience, which is unstructured, like observing, speaking, and writing. But after the emergence of computers, many of these data turned into structured and labeled data (emergence of databases adapted by ontology in philosophy).

In a world where Mid Journey generates digital art, Deep Mind AlphaFold predicts all possible protein forms, and WellSaid Labs has developed new voice actors for video games and other applications that can read phrases with their own expressions and pronunciations, we are in an age of intelligence. So, what would be the future of the web and services concerning these revolutionary changes?

Services with a specific's intelligence domain

In the index ventures summit, Sam Altman and Kevin Scott predicted that in 2022 we would see immense applications for what people can do with these models. Mid journey, the artificial intelligence program, was used by the British magazine The Economist to create the front cover for an issue in June 2022.

Despite these achievements, most AI-based services specialize in a specific intelligence domain, which means that there would be no general AI, and humans have yet to develop real intelligence. Several experts predict that microservices will be intelligent enough to satisfy a specific domain within two years (for example, answering questions about a particular product).

Many different architectures have seen the light of day since the 2017 work by Vaswani et al. (Paper, Attention is all you Need), which changed the nature of sequence-to-sequence models. Since then, many SOTA (state-of-the-art) language models have been developed.

API-based pre-trained models

When we look at hugging face ( An AI community that contains open datasets and pre-trained models ), we could predict that these models can expand to provide API-based services. Currently, there is a library named transformers that use these pre-trained models. Also, other platforms like Kaggle offer these datasets and notebooks and charge users based on GPU usage. Generally, many companies try to supply the infrastructure for the development of AI models. Google also provides Research Collaboration for offering GPU to training programs.

Using cloud-based resources provided by giant tech companies, we will probably see more API-based platforms, and integration and fine-tuning will become easier. Artificial Intelligence, in general, would become more applicable in platforms; as we have seen, in searching services like Algolia, no one needs to develop their search platforms. And, it would be competent for SME companies to know how to integrate these services.

The web would become much broader

Using cloud-based resources provided by giant tech companies, we will probably see more API-based platforms, and integration and fine-tuning will become easier. Artificial Intelligence, in general, would become more applicable in platforms; as we have seen, in searching services like Algolia, no one needs to develop their search platforms. And, it would be competent for SME companies to know how to integrate these services.